Tuning riboflavin derivatives for photodynamic inactivation of pathogens

The development of effective pathogen reduction strategies is required due to the rise in antibiotic-resistant bacteria and zoonotic viral pandemics. Photodynamic inactivation (PDI) of bacteria and viruses is a potent reduction strategy that bypasses typical resistance mechanisms. Naturally occurring riboflavin has been widely used in PDI applications due to efficient light-induced reactive oxygen species (ROS) release. By rational design of its core structure to alter (photo)physical properties, we obtained derivatives capable of outperforming riboflavin’s visible light-induced PDI against E. coli and a SARS-CoV-2 surrogate, revealing functional group dependency for each pathogen. Bacterial PDI was influenced mainly by guanidino substitution, whereas viral PDI increased through bromination of the flavin. These observations were related to enhanced uptake and ROS-specific nucleic acid cleavage mechanisms. Trends in the derivatives’ toxicity towards human fibroblast cells were also investigated to assess viable therapeutic derivatives and help guide further design of PDI agents to combat pathogenic organisms.


Results
Synthesis and characterisation of flavins. Flavins F1-4 (2A) were prepared according to Schemes S1 and S2 (see ESI). Inspired by the classical amphiflavins developed by Trissel, Schmidt and Hemmerich 31-33 , we chose to include an alkyl chain (C 8 ) within the structure to improve phospholipid membrane incorporation. Additionally, the flavin chromophore itself was substituted with bromo groups in the case of F3-4 in order to harness the heavy-atom effect that increases the rate of ISC from singlet to triplet excited states, thereby potentially enhancing photosensitised generation of singlet oxygen 34 . Heavy atom substitution has been shown to improve both the rate of ISC and the singlet oxygen quantum yield (Ф Δ ) in flavin derivatives used for synthetic photooxidation reactions 35 . Although the heavy-atom effect has been used to boost photodynamic efficacy for other PS dyes in PDI applications, it has not yet been explored for flavin derivatives 36 .
To afford both methylated (F1-2) and brominated derivatives (F3-4), a Boc-protected ethylene amino component was first installed to the methylated or brominated arene core prior to cyclisation of the isoalloxazine ring system. Following N 3 -alkylation with an octyl chain on the isoalloxazine, N-Boc protecting groups were removed to yield the amino-functionalised F1 or F3 that were subsequently converted to guanidino moieties to achieve F2 or F4 respectively (Fig. 2a). Guanidino groups were chosen as they are known to increase membrane coordination and penetration through strong guanidinium-phosphate H-bonding and we hypothesised would enhance PDI efficacy.
The UV-Vis absorption spectra of the flavins in DMSO revealed very similar absorption properties at the λ 1 (S 0 → S 1 ) band but a blue shift of 10-13 nm for the higher energy λ 2 (S 0 → S 2 ) bands of the brominated compounds (F3 and F4) presumably due to the electron withdrawing effect of Br atoms (Fig. 2b, Table 1, Table S1). In terms of emission properties, the heavy-atom effect of bromination can be clearly observed with F3 and F4 exhibiting severely reduced emission intensity (Ф F < 2% in DMSO, Table 1) when compared to the methylated F1 and F2 (Ф F = 13% and 14% respectively in DMSO, Table 1). This effect can also be observed when comparing the efficiency of 1 O 2 production upon excitation, where brominated F3 and F4 demonstrate up to a 30% increase in activity over the methylated F1 and F2 in MeCN (Table 1), indicating their potential to be potent photodynamic agents. Despite this, F1 and F2 are still efficient 1 O 2 photosensitisers comparable with riboflavin (Ф Δ = 0.54 ± 0.07). However, a smaller Ф Δ value for guanidinylated F4 (62%) is observed compared to aminated F3 (85%) which could be explained by fast reverse ISC and/or solvent-dependent aggregation.
After successful synthesis and initial characterisation, we monitored the photostability of the compounds in PBS (1x, pH 7.4) which was used for PDI assays as distilled or ultrapure water destabilises bacterial cells and coronaviruses through osmotic pressures 37,38 , thereby augmenting inactivation results. Flavins F1-4 and riboflavin (100 µM in PBS) were therefore irradiated with a 6200 K white LED light source (18 W, 400-700 nm, see ESI Figure S1 for emission spectrum) at an illuminance of 1 × 10 5 lx (35 mW/cm 2 irradiance) to resemble typical daylight 39 , and the changes in their UV-Vis absorption were monitored over time (see ESI, Figure S4). For riboflavin, rapid photodecomposition (80%) was observed over 30 min irradiation which is known to be due to intramolecular dealkylation of the ribityl chain ( Figure S4a,b). This yields lumichrome as the major degradation product which can only act as a PS under UV-irradiation 40 . Amino-containing F1 and F3 exhibited around 40% and 60% degradation respectively after the same irradiation time ( Figure S4a,c,e), whereas F2 and F4 degraded by approximately 10% and 30% respectively ( Figure S4a,d,f.). These findings corroborate previous work showing photodegradation of amino-containing flavins in the presence of phosphate ions, which resulted in diminished bacterial PDI efficiency 41 . However, it appears that guanidino substitution in our case improves the photostability. Interestingly, bromo-substituted flavins show higher rates of photodegradation, which could be explained by their higher Ф Δ values resulting in greater 1 O 2 -induced degradation.
Photodynamic inactivation of E. coli. Having observed clear trends in the photophysical properties of methylated and brominated derivatives, we were interested to see how this would affect the PDI efficacy of F1-4 against pathogens. First, we investigated the inactivation of the Gram-negative bacterium, E. coli BL21(DE3). The cell envelope of Gram-negative bacteria presents a formidable barrier to antimicrobial compounds that consequently inhibits PDI efficacy compared to the analogous structure in Gram-positive bacteria 42 . Following an initial 20 min incubation of the flavin compounds with E. coli in PBS at various concentrations in the dark, the mixture was irradiated (1 × 10 5 lx) and the number of surviving colony forming units (CFUs) were determined. After 15 min (31.5 J/cm 2 light dose) of irradiation, no inactivation was observed at 1 µM for either riboflavin or F1-4. At 10 µM approximately 1 log 10 reduction of E. coli CFUs was observed for F2 and a 2.8 log 10 reduction in the presence of F4 (Fig. 3a, Table S2). At higher concentrations, this effect was greatly enhanced with both F2 Table 1. Photophysical properties of flavins F1-4. λ n = S 0 → S n absorption band, λ em = emission wavelength, φ F = fluorescence quantum yield, φ Δ = singlet oxygen quantum yield. a measured in DMSO. b calculated using riboflavin as the reference (ϕ F = 0.226 ± 0.001 in DMSO). c calculated using Ru(bpy) 3 2+ as the reference (ϕ Δ = 0.57 ± 0.06 in MeCN). www.nature.com/scientificreports/ and F4 exhibiting > 6.0 log 10 reduction of CFU/mL at 100 µM which is considered a highest level of decontamination according to EMA guidance 43 .
Interestingly, riboflavin and F1 demonstrated no activity at 100 µM, whereas F3 exhibited a 1.8 log 10 reduction in bacterial load. To ensure effective PDI was occurring, the flavin compounds were incubated in the dark under the same experimental irradiation conditions (100 µM, 15 min) to reveal no bacterial toxicity (see ESI, Figure S5). The rate of E. coli inactivation over time was then monitored at a 100 µM flavin concentration revealing an extremely rapid reduction of CFUs in the presence of F4, facilitating > 6.0 log 10 reduction after just 5 min of irradiation (10.5 J/cm 2 light dose). Similarly, F2 shows effective bactericidal activity with a 3.4 log 10 reduction (> 99.9%) after 5 min irradiation. A closer investigation into the speed of F4's activity at 100 µM showed that after only 1 min of irradiation (2.10 J/cm 2 light dose) a 4.1 log 10 reduction (> 99.99%) of bacteria was achieved (Fig. 3b).
These data clearly show that the introduction of a guanidino moiety, as in the case of F2 and F4, greatly increases the flavin's PDI efficacy against the bacterium when compared to amino-containing F1 and F3. This may be attributed to improved photostability in PBS, as well as the guanidino group facilitating better coordination to the cytoplasmic phospholipid membrane resulting in improved permeability and lipid peroxidation under irradiation 44 . In addition, the replacement of methyl substituents with bromines improves efficacy when the same cationic group is compared. This can be explained by more efficient generation of singlet oxygen, as predicted by their Ф Δ values shown in Table 1. As a result, greater lipid and biomolecule oxidation can be achieved to inactivate the pathogen. It should also be noted that the inclusion of bromine atoms increases the lipophilicity of the molecule which can further improve cell membrane permeability and incorporation. However, despite having the highest predicted Ф Δ value (85%), F3 did not outperform F2 (Ф Δ = 55%) which demonstrates the overarching impact of guanidino substitution.
To gain initial mechanistic insight, we investigated the cellular localisation of fluorescent riboflavin, F1 and F2 by structured illumination microscopy (SIM) (Fig. 3c, see ESI Figures    . Although the exact quantification of flavin uptake was not possible due to uncharacterised optical properties in a complex biological environment, the qualitative comparison of F2 and F1 (which have similar Ф F in DMSO), revealed a higher number of fluorescent bacteria with bright fluorescent intensity after incubation with F2, especially within membranes (Fig. 3c, Figure S6 and S7). In addition, two subpopulations of bacteria can be observed, one in which intense flavin fluorescence is seen in the cell envelope and in the cytosolic space, and one in which fluorescence is seen only in the cell envelope and not in the cytosolic space. We believe that the existence of these two subpopulations indicate a level of cytosolic uptake in some cells. Although bacterial membranes serve as an excellent barrier for uptake of extracellular compounds, it is plausible that produced singlet oxygen might damage the integrity of bacterial envelope easing the penetration of flavin compounds. It can also be observed that guanidino substitution seems to lead to an enhanced uptake in comparison with other compounds. However, the distribution of fluorescent intensity was not homogeneous across different bacterial cells which could be related to the amphiphilicity of F2, resulting in variable uptake due to aggregation ( Figure S6). For hydrophilic riboflavin, very weak fluorescent populations were observed, most likely explained by the controlled transport of the compound through outer membrane porins of the bacterium and therefore unable to bind effectively to the outer or cytoplasmic membranes (Fig. 3c, Figure S8). This therefore helps to explain the lack of PDI efficacy observed when using riboflavin. Enhanced PDI efficacy of flavins that seem to be taken up into the cytosol can be explained by their ability to interact with biomolecules within the cell, such as nucleic acid. To test this hypothesis, we monitored the pDNA (pUC18) cleavage in the presence of the flavins (10 μM) under irradiation, which indicated guanidino derivatives possess superior photocleavage ability (Fig. 3d). The photocleavage of supercoiled (SC) to nicked coiled (NC) pDNA structures was significantly enhanced for F2 and F4 (39% and 21% respectively) when compared to F1 and F3 (≤ 3%) after irradiation for 15 min (Fig. 3d). This difference in activity could be explained by favourable guanidinium-phosphate interactions that increase the likelihood of electron transfer events between flavin and guanosine, which are known to primarily contribute to DNA cleavage alongside 1 O 2 -mediated oxidation 45,46 . As a consequence, even if amino-flavin compounds were able to diffuse into the cytoplasm, it is unlikely that they will cause damage of the constituent nucleic acids. Our control compound, riboflavin also exhibits photocleavage of the plasmid (11%) which has been reported previously 45,46 . However our SIM data indicate that, due to its hydrophilic nature and lack of cationic substituent, riboflavin does not bind readily to bacterial cells resulting in limited PDI.
Collectively, this initial mechanistic study demonstrates that even in the case of increased singlet oxygen production, the key component to achieve effective Gram-negative bactericidal activity under irradiation is the presence of the guanidino group, which facilitates cell uptake and nucleic acid degradation. It should be noted, that although considered to be a powerful super resolution cross-sectional microscopy, SIM alone cannot be used as a conclusive evidence for cytoplasmic and additional strategies should be employed. However, the combination of distinct populations of bacteria in presence of different flavins in combination with enhanced DNA cleavage of guanidino compounds, speaks of at least some level of uptake.

Photodynamic inactivation of murine hepatitis virus (MHV-A59).
Encouraged by the identification of such highly effective guanidino-flavins for bacterial inactivation, we were interested to see if the trend would be similar for coronaviruses, specifically, the coronavirus surrogate, murine hepatitis virus A59 strain (MHV-A59) 30 . It has already been demonstrated that riboflavin can effectively inactivate both enveloped and non-enveloped viruses in blood products using UV light [21][22][23] , but much lower efficacy was observed using visible light (0.4 × 10 5 lx, 0.5-2 h) against hepatitis B virus (HBV) 47,48 . In order to evaluate possible applicability towards virus-inactivating surface coatings or textiles, we used an in vitro TCID 50 assay to evaluate the viral titre of MHV-A59 through inoculation into murine fibroblast 17Cl-1 cells after irradiation at varying concentrations of flavin in PBS. The cytotoxicity of the flavins towards this cell line was first investigated by 24 h incubation MTS assay allowing us to obtain a working concentration range of ≤ 10 μM for the in vitro TCID 50 assay to evaluate viral PDI efficacy (see ESI Figures S11 and Table S4 for 17Cl-1 cytotoxicity data).
We started our investigation by varying the concentration of flavin (1-10 μM in PBS) with 10 min of white LED exposure (21.0 J/cm 2 light dose, Fig. 4a). Even at 1 μM, the brominated guanidino flavin F4 demonstrated a 3.4 log 10 reduction in viral titre, while F2 and F3 showed 2.1 log 10 reduction (Fig. 4a, Table S3). It should be noted that reductions of the order of 4 logs or more (> 99.99%) are considered highly effective by EMA guidance and that a > 1 log 10 reduction in necessary to be considered reliable 49 . At the same concentration, riboflavin achieved a 1.2 log 10 reduction in titre, however F1 showed no effect under these conditions. Nevertheless, the activity of all flavins improved by the increase of their concentrations resulting in viral load reductions of > 99.9% in the presence of riboflavin or F2, and > 99.99% with F3 or F4 at 5 μM. Further increase in concentration (to 10 µM) only substantially improved virucidal activity for F4 (> 5 log 10 ). The irradiation time was then explored for 10 μM flavin to reveal high degrees of inactivation (≥ 5 log 10 ) for F2, F3 and F4 when irradiated for 15 min (31.5 J/cm 2 light dose) whereas shorter irradiation times (5 min, 10.5 J/cm 2 light dose) still provided efficient inactivation of MHV-A59 ≥ 3 log 10 steps using those same flavins (Fig. 4b). It should be noted that variability within all experiments was noticeably high which is typical for TCID 50 assays 50 . However, no reliable log reduction of MHV-A59 was observed when irradiated without flavin in PBS containing 0.1% DMSO (Fig. 4b). To confirm that the mechanism of viral inactivation by the flavins was dependent on light, dark control experiments (10 μM, 10 min incubation) resulted in no effective virucidal activity being observed (> 1 log 10 reduction, Fig. 4c).
The activity of the flavin derivatives towards viral inactivation show a different trend than previously observed with our model bacterium. For example, in the presence of riboflavin under the same conditions, effective PDI of MHV-A59 was measured whereas there was no activity towards E. coli. Therefore, it can be assumed that the requirements for a PS to interact and/or diffuse through the coronavirus membrane are less dependent on www.nature.com/scientificreports/ lipophilicity or the presence of cationic charge. Despite this, our flavin derivatives again show that guanidino group incorporation improves antiviral PDI activity when compared to amino groups. However, bromination now seems to play a more important role in viral deactivation as brominated amino F3 outperformed methylated guanidino F2.
To try and explain our findings, we investigated the interaction of the flavins with RNA to elucidate how its light-induced cleavage could lead to coronavirus inactivation, a model ssRNA (~ 1 k nt) was irradiated (1 × 10 5 lx) in the presence of 10 µM flavin in PBS for 15 min and the extent of cleavage was observed via agarose gel electrophoresis (Fig. 4d). It was observed that F1 exhibits the least RNA photocleavage (12%) whereas riboflavin shows greater activity (28%) which corroborates the result of viral inactivation and confirms previous findings [51][52][53] . Similar to our results with pDNA, guanidino-containing F2 and F4 show the best photocleavage ability (46% and 42% respectively). Interestingly, F3 also exhibits effective cleavage of RNA (35%) which was not observed in the case of pDNA (Fig. 3d). This can be rationalised by the mechanism of flavin-mediated RNA photocleavage which has been previously shown to depend more upon 1 O 2 oxidation than electron transfer events between flavin and nucleobase [51][52][53] . Taken together, it can be concluded that efficient 1 O 2 production is key to viral PDI using flavin derivatives.
In vitro toxicity towards human cells. Finally, before considering the applications of our flavin derivatives for PDI of topical pathogen infections, surface coatings or textiles, we investigated their impact on human cells. In general, an ideal photosensitiser for these applications should show no cytotoxic effects on human cells in the dark and limited effects given the same irradiation conditions used for inactivation of the target pathogen, also referred to as a therapeutic window 12 . Accordingly, the inherent cytotoxicity of the flavins was evaluated through incubation with human lung fibroblast cells (WI-38). A 24 h MTS assay was used to calculate inhibitory concentration values (IC 50 ) of > 100 µM for riboflavin and guanidino-functionalised F2 and F4 which can therefore be considered as non-toxic (Fig. 5a, Table S4). On the other hand, amino-functionalised F1 and F3 had IC 50 values of 96.1 µM and 30.9 µM respectively (Table S4). It has been shown previously that the cytotoxicity www.nature.com/scientificreports/ of amino-containing compounds may be derived from an increase in intracellular amine oxidase activity that induces excess oxidative stress leading to cell apoptosis 54 .
We then studied the light-induced toxicity of the flavins towards the WI-38 fibroblast cells by observing decreases in cell viability over irradiation time using an MTS assay (Fig. 5b). Interestingly, the cytotoxicity trends observed in the dark after 24 h are reversed when exposed to light (1 × 10 5 lx) over shorter time periods. For example, after 15 min of irradiation, 10 μM of methylated guanidino F2 induces ~ 75% reduction in cell viability. Under the same conditions, brominated guanidino F4 exhibits less of an effect with ~ 20% reduction in cell viability. The degree of variability in these experiments could be attributed to the amphiphilic nature of the flavins leading to aggregation of the compounds in aqueous conditions. On the other hand, riboflavin, amino-functionalised F1 and F3 display no decrease in cell viability over 15 min irradiation. These data show that despite being non-toxic to the cells in the dark, the guanidino-functionalised flavins can induce cytotoxicity upon irradiation most likely through similar mechanisms discussed for pathogens. It is our ongoing work to understand the differences in light induced cytotoxicity that were observed between methylated and brominated guanidino flavin derivatives. Nevertheless, F4 could be suitable for further PDI applications thanks to effective bacterial and viral inactivation (> 3 log 10 reduction after 15 min irradiation at 11 μM and 1 μM respectively) coupled with low toxicity to human fibroblast cells under the same conditions or in the dark.

Conclusion
In summary, we have rationally designed small library of flavin derivatives containing functional groups that significantly improve visible light photodynamic inactivation of pathogens. By incorporating a guanidino moiety into the flavin structure (F2 and F4), the inactivation efficacy against Gram-negative bacteria, E. coli BL21(DE3) in PBS, was remarkably enhanced compared to natural riboflavin and amino variants (F1 and F3). We believe that this can be explained by enhanced cell uptake of these compounds and subsequent pDNA cleavage. Bromination of the structure (F3 and F4) enhanced singlet oxygen production via the heavy-atom effect which generally led to improved PDI of E. coli, however the incorporation of a guanidino substituent dominated the outcome. Very effective inactivation was achieved with F4, whereby 1 min of visible irradiation (2.10 J/cm 2 light dose) reduced the bacterial load by > 4 log 10 steps.
It was found that the presence of guanidino group does not enhance the selectivity of the compounds in case of PDI of SARSCoV-2 surrogate, MHV-A59 when evaluating the PDI against the SARS-CoV-2 surrogate, MHV-A59 in PBS. Rather, the bigger effect is exerted by presence of bromo substituents; brominated amino F3 had a higher effiecenty compared to methylated guanidino F2 compound, which can be attributed to more significant 1 O 2 -mediated RNA cleavage. This was evidenced by brominated amino F3 having a higher efficacy compared to methylated guanidino F2 which was attributed to greater RNA cleavage via 1 O 2 oxidation. The highest viral PDI was achieved by F4 where > 4 log 10 reduction in titre was achieved after 10 min of visible irradiation (21.0 J/ cm 2 light dose) in the presence of 5 μM compound. Toxicity studies using the novel flavins towards human fibroblast cells (WI-38) further confirmed that F4 could be a suitable candidate for PDI applications against both bacteria and viruses.
Overall, we hope this study inspires further exploration of riboflavin's structure to design new generations of photosensitiser compounds to treat topical bacterial infections, enabling the design of antiviral surface coatings, sprayable hybrid materials to provide the surface protection, as well as fabrics capable of efficient visible light pathogen inactivation. Such fabrics could be used to manufacture protecting suits for medical staff, as well as curtains and bedding used in intensive care units where the sterile conditions are of the outmost importance.   After an overnight aerobic culture in LB broth at 37 °C with shaking at 225 rpm, the resulting E. coli were centrifuged at 2,500 rpm for 10 min. The pellet was resuspended in PBS (Sigma, 1x, pH 7.4) to an OD 600 value of 0.6 (~ 1 × 10 8 bacteria per ml). Bacterial solutions were then combined with varying concentrations of F1-4 and riboflavin (0, 1, 10, 100 μM) dissolved in PBS in a 1:1 ratio and left to incubate for 20 min in dark. Aliquots of negative controls (0.5% DMSO in PBS; PBS only) and 100 μM flavin-bacteria samples were covered in foil to remain in dark, while all samples were plated onto a 96-well plate and exposed to white LED irradiation (1 × 10 5 lx) for 15 min. After 15 min of irradiation or darkness, each sample was serially diluted to its respective, appropriate dilution and plated onto an LB agar plate with 50 μg/mL kanamycin. Plates were incubated overnight at 37 °C in the dark. Survival of bacteria was determined by counting colony forming units (CFUs) the next day.

Methods
For time point data, aliquots of 100 μM flavin-bacteria solutions were taken at 5-min increments of irradiation (0, 5, 10, and 15 min) and evaluated similarly as above. Time point experiment with one-minute increments, between 0 and 5 min, were conducted for 100 μM F4-bacteria samples.
All data was generated with technical replicates and biological triplicates. All statistical analysis was done with Graphpad Prism 9. One-way ANOVA with Tukey's multiple comparisons test, was performed to evaluate the impact of flavin concentration compared to the control (PBS). Two-way ANOVA with Bonferroni's multiple comparison test, was performed to evaluate the impact of light given the same flavin concentration with different irradiation conditions (dark and light). Significance levels are defined as the following: ns for p > 0.05, * for p ≤ 0.05, ** for p ≤ 0.01, *** for p < 0.001, and **** for p < 0.0001.
Virus propagation. MHV-A59 was propagated in 17Cl-1 cells at a multiplicity of infection (MOI) of 0.01 TCID 50 per cell. MHV-A59 was clarified by centrifugation at 3000 rpm for 10 min and stored in aliquots at − 80 °C. The TCID 50 was calculated using the Reed and Muench method 55 .
Flavin treatment of MHV-A59. The efficacy of the flavins were enumerated utilizing modified methods from ISO 18,184 and Leibowitz et al. 56 Flavins F1-4 and riboflavin, were solubilized in DMSO. 100 µM working stock solutions were then created by diluting each DMSO solution into PBS. To control for the effect of DMSO on MHV-A59, DMSO was diluted into PBS at the same concentration. A white light apparatus was set up to cover a 4 × 5 (row x column) area of a 96-well plate with the same level of light intensity (1 × 10 5 lx). The stock solutions were then diluted in replicates of four to a total of 180 µl in PBS at 1, 5, and 10 µM concentrations for each sample. Virus stocks of MHV-A59 were thawed on ice prior to use. 20 µl of MHV-A59 stock (1.4 × 10 9 PFU/ml) was then mixed into each well, and the 4 × 5 well matrix was treated with the light or incubated in the dark at room temperature for a specified time (5, 10, or 15 min). 50 Assay MHV-A59. After treatment, the samples were mixed into 320 µl of Gibco Difco™ Beef Extract (1.5% w/v beef extract in ddH 2 O; Life Technologies) and rolled for 15 min to chelate any free ions in the solution. The samples then were serially diluted (10 0 -10 -7 ) in infectivity media (Dulbecco's modified Eagle's medium low glucose 1 g/L (DMEM, Life Technologies) supplemented with 2.5% fetal bovine serum (Merck), 3% tryptose phosphate broth (Merck), 1 × non-essential amino acids (Gibco), 1 × antibiotic-antimycotic (Thermo Fisher Scientific) and 1 × L-Glutamine (Gibco]) and added in triplicate to 17Cl-1 cell plates. For each sample or control there were 4 replicate treatments (in light or dark); each treatment replicate was plated in triplicate. Infected 17Cl-1 plates were incubated at 37 °C in a 5% CO 2 atmosphere for 24 h. Plates were scored for cytopathic effect (CPE) by microscopy and viral titres were determined by the Reed and Muench 50% tissue culture infectious dose (TCID 50 ) end point method 55 . Outliers were removed from the four replicates using an outlier test. The full protocol was repeated in triplicate on different days to account for any variability in the assay.

Virus recovery and TCID
Nucleic acid photocleavage. General procedure: Riboflavin or flavin F1-4 (10 µM from DMSO stock solutions) were mixed with either pUC18 DNA (Thermo Scientific, 0.50 µg) or EGFP-encoding mRNA (~ 1 k nt) as the ssRNA model (TriLink Biotech, 0.88 µg) dissolved in PBS (40 µL) to give a final DMSO concentration of 0.1%. A 20 µL aliquot was taken after 15 min of white LED irradiation (1 × 10 5 lx). Prior to loading, 2 µL of 6X Orange-G loading dye in glycerol were added to the aliquot and mixed vigorously. The gel was electrophoresed at 80 V for 35 min with 1% agarose gels in 1X Tris-acetate-EDTA (TAE) buffer. Gels were then imaged in a Syngene G:BOX Gel Documentation System and band % quantification was performed using ImageJ Gel Analyzer.
MTS viability assay. The effect on cell viability of WI-38 and 17Cl-1 cells after treatment with F1-F4 and riboflavin was determined using the commercially available MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H tetrazolium] assay (Promega). The MTS tetrazolium compound is reduced by cells into a coloured formazan product which is soluble in cell culture media. It can be detected colorimetrically between 450 and 540 nm with the measured absorbance directly proportional to the amount of metabolically active cells in culture. Cells were seeded into clear 96-well plates containing 10,000 cells/well in 100 µL complete growth medium and cultured for 24 h at 37 °C and 5% CO 2 . Subsequently, cells were treated with varying concentrations of F1-F4 and riboflavin (0.01-100 μM) dissolved in complete growth media containing 0.1% DMSO. After further 24 h incubation at 37 °C and 5% CO 2 , 20 µL of CelTiter 96® AQ ueous One Solution (Promega) was added into each well and incubated at 37 °C, 5% CO 2 for 1-4 h, according to the manufacturer's instruction. The absorbance of each well was measured at 490 nm using a plate reader (Spark, Tecan). Control measurements included negative control of cells with DMEM, cells with DMEM containing 0.1% DMSO, cell-free culture media (blank) and cell-free sample dilutions in culture media to evaluate potential sample interferences with MTS assay. All experiments were conducted in biological triplicates. The percentage cell viability was calculated according to Eq. (1): An adapted method reported by Maisch et al. was used to conduct phototoxicity studies on human lung fibroblasts, WI-38 28 . Cells were seeded into 96-well plates containing 10,000 cells/well in 100 µL complete growth medium and cultured for 24 h at 37 °C and 5% CO 2 . On the next day, the growth media was removed, and cells were treated with 10 μM F1-F4 and riboflavin dissolved in DMEM without serum and phenol red (Gibco) containing 0.1% DMSO. The resulting mixtures were then incubated in the dark for 5 min and then either illuminated with 1 × 10 5 lx for 5, 10 or 15 min, or incubated further in the dark for 15 min (dark control). After irradiation, the flavin solutions were removed and 100 μL of DMEM with 10% FBS and without phenol red was added to each well and incubated over night at 37 °C and 5% CO 2 . Subsequently, 20 µL of MTS reagent was added into each well and incubated at 37 °C and 5% CO 2 for 1-4 h and the absorbance of each well was measured at 490 nm using a plate reader. Control measurements included negative control untreated cells with light and in the dark containing 0.1% DMSO and cells treated with flavins for 15 min in the dark. All experiments were conducted in biological triplicates. The percentage cell viability was calculated according to Eq. (1).

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.